Principles for Using AI Agents to Keep Your Commitments
As organizations plan for 2027, many are deciding what work AI agents will take on next year.
Successful businesses take ownership of their obligations and make commitments to meet them.
These commitments are promises that must be kept. That is the lens worth bringing to AI:
how can AI agents help us keep the promises we have made?
Obligations and promises
Over the last year in our Elevate Compliance Huddles, we have walked through the Operationalized Obligation model, which is based on Promise Theory. It describes how obligations become promises.
An obligation begins with an intention. Once imposed, it creates uncertainty about how and whether it will be met. A promise answers that uncertainty. It states what will be done and how it will be done, and the promise owner commits to doing the right work, the right way, to achieve the right results. That commitment is what brings certainty.
Operational compliance is the work between the obligation and the promise. It is where promises are kept, obligations are fulfilled, and evidence of compliance is produced. Each side of the model has its own work:
Obligation owners manage what is required through rules, practices, targets, and outcomes. They audit and validate that obligations are being met.
Promise owners keep their promises through processes, systems, programs, and governance. They assess and verify that those promises are being kept.
When we ask an agent to help
A promise owner can be a person or a machine. People make promises through their commitments. A machine makes promises through its design, and it can only keep the promises it was designed to keep.
When we ask an AI agent to help, we are transferring responsibility for keeping part of a promise we have made. Accountability stays with the obligation owner, who answers for the work the agent does and the results it achieves. The left side of the model remains with them: knowing what was intended and validating that obligations are being met.
Before an agent is given a promise to keep, the questions are the same ones we would ask of anyone we rely on:
Does it have the knowledge, know-how, and skills the work requires?
Does it have the capacity to keep the promise every time it is asked, under the conditions it will actually face?
An agent with the right capability and capacity extends what the business can deliver, and knowing whether it has them is where the work begins.
Principles for AI adoption
All of this points to a more fundamental question: what principles are we applying to AI adoption?
Most AI adoption is not engineered. It proceeds without the management, compliance, safety, and other principles we apply to the rest of the business. The Operationalized Obligation model points to principles that are relevant here:
Ownership. Every obligation must have an owner who is accountable for meeting it.
Promises by design. A machine makes promises through its design, so its design must reflect the commitments the business has made.
Retained accountability. Responsibility may be transferred to an agent. Accountability must remain with the obligation owner.
Capability before transfer. No agent should be asked to keep a promise until its capability and capacity have been established.
Verification and validation. Promises must be verified as kept, and obligations validated as met.
Looking ahead to 2027
As the year ahead takes shape, these questions are worth asking:
What principles are guiding your adoption of AI?
Which of your commitments could AI agents help you keep?
How will you know those agents are capable of keeping them?
If you are working through these questions too, I would love to hear from you.
Raimund Laqua, PMP, P.Eng.
Founder, Principal Consultant
Lean Compliance Consulting, Inc.
Engineering Trust in the Intelligence Age
